Open Nutrition Dataset
Nutrition datasets form an important foundation for nutrition counseling, disease prevention, and medical dietetics, including the management of chronic diseases. However, many nutrition datasets are inconsistent.
Nutritional databases provide a scientifically sound basis for determining nutrient requirements and deriving dietary recommendations. However, many databases are inconsistent. They are based on different recipes, focus heavily on specific regions — such as US cuisine — or classify items too broadly — 'a serving of pasta' instead of '100 grams of pasta'.
This is where we come in. We are developing an open dataset for scientific use focusing on Central European, and specifically German, cuisine. Each ingredient is described using a consistent ID, and synonyms such as 'leek' and 'spring onion' are standardised. The edible portion and whether an ingredient is raw or cooked are accurately recorded, and the weight of each ingredient in a recipe is stored in a uniform schema. These documents provide a systematic evaluation and comparison of the ingredients used in a dish. We also store high-resolution images of each dish across multiple preparation stages, perspectives and shooting conditions to support visual analysis methods.
Once the ingredients are accurately recorded and defined, the macro- and micronutrient values—such as carbohydrates, fats, proteins, vitamins, and minerals—can be reliably determined for each dish. To do this, we link our nutritional dataset to the Bundeslebensmittelschlüssel (BLS), the official German nutritional database.